Today's curated cutting-edge AI news, welcome to read 👇
🎵 The Doubao team releases the music generation framework Seed-Music, comparable to Suno, supporting timbre cloning and fine-grained style control!
🚀 OpenAI increases the usage limit of the o1 model, significantly boosting the weekly usage for Plus users!
🎥 Luma AI launches AI Video API, the first publicly available AI video product!
🖌️ InstantDrag technology framework, allowing direct image editing through dragging, generating effects interactively within 1 second!
🎨 OmniGen unified image generation model, achieving conditional control for drawing without additional models, looking forward to open-source release!
Cutting-edge News
1. Doubao Team Releases Music Generation Framework: Seed-Music, Comparable to Suno!
In addition to supporting song generation from lyrics, converting lyrics to sheet music and then to songs, and song editing, there are two additional highlights:
- It supports cloning the timbre of the original song and generating a new song, directly productizing the previously popular AI Stefanie Sun project.
- It supports fine-grained style control, allowing the melody to remain unchanged while only modifying the lyrics.
Detailed introduction: https://team.doubao.com/en/special/seed-music
Currently, no public trial is available, only a few demo presentations have been released. From what I’ve seen, it seems to be even stronger than Suno!
2. OpenAI increases the usage limit of the o1 model.
On the 17th, it announced that for Plus and Team users, the usage of the o1 Preview model will increase from 30 times/week to 50 times/week, while the o1 mini model usage will increase from 50 times/week to 50 times/day.
3. Luma AI launches AI Video API.
It becomes the first publicly available API for AI video products. Before, companies like KooLing also launched API functions, but it required special application.
Website: https://lumalabs.ai/dream-machine/api
Cutting-edge Technology
1. A technology framework for image editing via dragging: InstantDrag.
Directly drag the parts and directions you want to modify in the image, no need for input prompts, and the generated effect appears within 1 second for interactive image editing.
Detailed introduction: https://joonghyuk.com/instantdrag-web/
AI Drawing
1. A unified image generation model: OmniGen.
Aims to solve multiple image generation tasks with one model, achieving drawing condition control without the need for additional models like ControlNet or IPAdapter.
GitHub: https://github.com/VectorSpaceLab/OmniGen
Paper: https://arxiv.org/pdf/2409.11340
Currently, only the paper has been released, with plans for an open-source release later.
